Malcolm S. Longair
Cavendish Laboratory, University of Cambridge, Cambridge CB3 0HE,
England.
, the co-moving space density of
star-forming galaxies is approximately half of that of luminous galaxies
(
) at the present day. Radio galaxies at redshifts 
have high optical surface brightness emission regions stimulated by the
passage of the radio jets responsible for powering their large-scale
radio structures.
Keywords: galaxies, cosmology, galactic evolution, radio galaxies
It is impossible to do justice in a modest space to the vast amount of new information which has resulted from Hubble Space Telescope observations of distant galaxies. I will touch briefly on the following areas which will be dealt with in much more detail in the contributed papers and in the posters:
.
It is important to appreciate exactly what it is that we observe when
optical HST observations extend to redshifts 
. In fact, the
predictions are not so different from those presented by Gunn (1979).
For example, Figure 1(a) shows the nearby giant spiral galaxy NGC 5364
at a redshift of 0.004137 and Figure 1(b) the same galaxy as it
would appear in an 8000 second observation through the I filter, F814W,
if it had redshift 
(Abraham et al. 1996). It can be
seen that the spiral structure is still just detectable but, at greater
redshifts, it would be impossible to distinguish details of its
morphological structure.
Figure: (a) NGC 5364 as observed from the ground; (b) NCC 5364 as it
would be observed at a redshift 
by the WFPC2 in an 8000 second
integration through the I filter, F814W (Abraham et al.
1996).
The above simulation was made using estimates of the K-corrections for spiral galaxies. Giavalisco et al. (1996) have described an improved procedure using Astro-1 Shuttle observations of nearby galaxies in ultraviolet wavebands at 152 nm and 249 nm. The galaxies M32, M31, M33, M74, M77, M81 and M82 were observed in these wavebands by the Ultraviolet Imaging Telescope of the Astro-1 mission and Giavalisco and his colleagues have used these to determine the morphologies and apparent magnitudes these galaxies would have if observed by the WFPC2 in 4.3 hour integrations. Some of their results are summarised in Table 1.
The figures enclosed in square brackets in Table 1 indicate that, if the
galaxies were located at these redshifts, they would not be observable
by the HST because they have too low surface brightnesses. NGC
1068 is a special case since it possesses a Seyfert nucleus and so
can be observed even at the very large redshift z = 3, but the host
galaxy would be undetectable. The two other galaxies which are just
detectable at redshift 
are observed with too low
signal-to-noise ratio for them to be classified morphologically. These
computations make the important point that, if galaxies are to be
classified successfully at redshifts z > 1 by the HST, they must
either have greater surface brightnesses or luminosities than the
typical galaxies we observe nearby. We will find that there are good
observational reasons why this is, in fact, the case. It is also clear
that it is possible to make excellent progress in the morphological
classification of galaxies out to redshifts of about 0.5, and greater
for the more luminous galaxies.
Table: Local galaxies observed by the Ultraviolet Imaging Telescope
of the Astro-1 Shuttle Mission. The table shows the expected magnitudes
of the galaxies if they were observed at redshifts 
and 
by the HST. Square brackets indicate that the galaxies would be of
too low surface brightness to be detected (Giavalisco et al.
1996).
Several studies of the morphologies of faint galaxies have been
reported (Glazebrook et al. 1995a, Abraham et al. 1996,
Driver et al. 1995,1996). Different approaches have been taken to
the assignment of morphological types. The validation of computer
algorithms to assign types is of importance in order to
determine the morphological mix of the faintest galaxy samples. In the
approach taken by Abraham et al. (1996), morphological types
have been assigned by galaxy experts and these have been compared with
`objective' machine classifications based upon two parameters which can
be conveniently defined for each galaxy image. These are the compactness
of the brightness distribution
within the image of the galaxy, C, and its asymmetry, A, determined
by rotating the galaxy image through 180
and defining
objectively
its degree of asymmetry. These parameters have been
determined for a sample of galaxies from the HST Medium Deep Survey,
which has been carried out in parallel mode for a number of random
fields at high galactic latitudes. Typically, the exposures were for 5000
seconds and 507 objects were classified with I<22. The A-C diagram
is shown in Figure 2, in which the morphological types have been
assigned by van den Bergh according to a scheme in which compact,
elliptical and S0 galaxies are denoted by ellipses, spiral galaxies
of all types by spirals, and irregulars, peculiar and
merging systems by crosses. The diagonal lines on Figure 2
show the boundaries within which it is expected that the three different
morphological classes are found, on the basis of redshifting examples
of Hubble types to faint apparent magnitudes. According to Abraham
et al., the machine assignments agree very well with the `expert'
classifications down to apparent magnitude I = 21, but there is more
scatter in the assignments in the faintest magnitude interval, 21 <
I < 22.
Figure: The distribution of galaxies on the central concentration,
C, versus asymmetry, A, plane. The morphological assignments by
van den Bergh are shown according to a scheme in which compact galaxies,
ellipticals and S0 galaxies are denoted by ellipses, spiral galaxies of
all types are designated by spirals, and irregulars, peculiar and
merging systems are indicated by crosses. Representative error bars for
different regions of the diagram are shown. The regions of the plane
corresponding to the three morphological categories, as derived from an
artificially redshifted sample of local galaxies, are also shown (Abraham
et al. 1996.)
These procedures are of considerable importance for understanding the
excess counts of faint blue galaxies observed in deep galaxy surveys.
Figure 3 shows the counts of galaxies of different morphological types.
The solid lines show the expected counts for uniform world models with
and it can be observed that the E/S0 and spiral
samples follow closely these expectations. The important result is
that the excess blue galaxies are associated with the galaxies which
are classed as Irregulars/Mergers.
Figure: The counts of faint galaxies of different morphological
types. The counts as determined by van den Bergh, Ellis
and according to automatic classification procedures for the same
sample of galaxies are indicated by different symbols. The solid lines
show the expected galaxy counts for a uniform world model with no
evolution. The pronounced excess of irregular/merging galaxies is
apparent (Abraham et al. 1996).
A similar result has been found by Driver et al. (1995) in their
analysis of a single ultra-deep WFPC2 deep field which was observed for
5.7 hours in each of the V (F606W) and I (F814W) wavebands. These
samples extend somewhat deeper than the Medium Deep Survey, the limiting
apparent magnitudes being 
, corresponding roughly to 
. In their analysis, the classification of the images was carried
out by fitting surface brightness profiles to the galaxy images, as well
as by visual inspection. The ellipticals are those for which the
surface brightness distributions follow de Vaucouleurs' 
law
and the spirals those with exponential light profiles. Those with light
profiles which follow neither of these distributions are classified as
irregulars. They find exactly the same result as in the Medium Deep
Survey, but now the counts extend to significantly fainter apparent
magnitudes.
In an intriguing study, Neuschaefer et al. (1996) analysed the spatial
distribution of galaxies in 28 fields from the Medium Deep Survey and in
the Groth-Westphal survey, which consisted of 28 contiguous fields at
high galactic latitudes. Many close pairs and apparently interacting
systems of galaxies were noted among the blue population. One of the
most striking results of their analysis is the determination of the
two-point correlation function for galaxies down to the very smallest
angular separations. Despite the fact that there are apparently many
associations of faint galaxies in the samples, the two-point angular
correlation functions seem to follow the standard result 
down to angular scales as small as 2 arcsec,
with only a small excess of assocations at small angular radii.
Thus, although it might have been expected that there would be a large
excess of nearest neighbours if the blue galaxies were due to
interacting or coalescing systems, it seems that the very modest excess
does not greatly exceed what might be expected from the standard two-point
correlation function.
There has been considerable debate about the nature of the excess faint blue galaxy population. It had been expected that there might well be more blue galaxies at faint apparent magnitudes because even passively evolving models of galaxies suggest that the old stellar population should be brighter in the past and there should be more star formation activity. One of the surprises of the first redshift surveys, which extended to apparent magnitudes at which the blue excess was observed, was that the mean redshift of the galaxies at the faintest magnitudes did not increase any more rapidly than would have been expected if the galaxy population had remained unchanged with cosmic epoch (Glazebrook et al. 1995b). This result was interpreted as indicating that the excess blue galaxies were associated with a population of dwarf galaxies. There must, however, be more to the story than this.
It is now becoming possible to study fainter samples of galaxies and
three examples indicate what is now becoming the state of the art. At
this meeting, Schade reported the most recent results of the
Canada-France Redshift Survey (Le Fèvre et al. 1995). This
magnitude-limited complete survey contains 943 objects with magnitudes
. The objects for which redshifts have
been measured extend to redshifts 
. HST images have been
secured for 32 randomly selected galaxies with redshifts 
and
these display the normal range of morphological types. There are,
however, important differences. The mean rest frame surface brightnesses
of the late type galaxies are about 1.2 magnitudes greater than those of
nearby galaxies. Some degree of peculiarity/asymmetry is observed in
30% of the objects, and 13% show clear signs of mergers or
interactions. There are compact blue components in 30% of the galaxies
and these occur predominantly in the peculiar systems, but a few of them
are also present in normals systems. According to Schade et al.
(1995), these galaxies are predominantly the cause of the excess of
faint blue galaxies and the numbers are consistent with those found in
the morphological surveys described in Section 3.
The second example concerns the recent results of Cowie et al.
(1995) on the redshift distribution of faint galaxies. The galaxies
were selected from very deep surveys in small areas of sky and consisted
of all the objects in these areas which satisfied the magnitude
selection criteria K < 20, 
and 
. Spectra for these
galaxies have been obtained with the Keck 10-metre telescope and the
programme has been remarkably successful in discovering large redshift
galaxies. There were 367 objects which satisfied the selection criteria
and, among them, 91 are already known to be galaxies with redshifts 
and 40 with redshifts z > 1. The reason for their success is
immediately apparent from the typical spectra of the galaxies for which
spectra have been obtained (Figure 4). The large redshift galaxies have
strong emission lines, characteristic of regions of star formation.
According to their analysis, these are luminous galaxies with absolute B
magnitudes, 
. The luminosities of the large redshift
galaxies in the [OII] line are typically at least an order of magnitude
greater than those of a reference sample at redshifts 
. They
infer that these are star-forming massive galaxies at redshifts 
and
the rates at which the stars are being formed can account for between
about 5 and 20% of their present stellar populations.
Figure: The average rest-wavelength spectra of the objects in the
deep survey field SSA13. It is apparent that they are strong emission
line objects with only weak absorption lines in the ultraviolet region
of the spectrum (Cowie et al. 1995).
Deep HST images have been obtained for nine of the galaxies at redshifts z > 1 in the I waveband. As they note, the galaxies have `strikingly unusual morphologies, often consisting of chains or structures of compact blobs, suggesting that they are generally not dominated by uniformly distributed star formation'. These results suggest that among the faint blue galaxies there is a population of distant star forming galaxies in addition to objects at smaller redshifts.
Similar preliminary results were reported by Koo at this meeting from a
deep survey undertaken with the Keck 10-m telescope in the HST Groth
strip. The sample extends to I = 24 and in this area they already
have redshifts for 35 galaxies in the range 
, with a mean redshift of about 0.8, significantly greater than that
of the Canada-France Redshift Survey. Again, the sample contains a
large number of unusual galaxies including objects with multiple knots
and the types of chain galaxy reported by Cowie et al. (1995). In
addition, they have found nine red galaxies which are as red as elliptical
galaxies are at the present epoch. The inference is that these galaxies
have already had time to form old stellar populations and so must have
undergone their last major burst of star formation at redshifts z >
2.
Another beautiful example of the importance of mergers and interactions
between galaxies is provided by the very deep radio-optical survey of
Windhorst et al. (1995). They conducted a very deep VLA
survey of a small area of sky which reached a flux density limit of
about 1 
Jy. The corresponding deep optical survey carried out by
the HST extends to 
. It turns out that 60% of the
microjansky radio sources are identified in this survey, many of them
being associated with pairs or groups of galaxies. These are such
intrinsically weak radio sources that they are not much more luminous as
radio sources than normal galaxies. It is natural to associate them
with the types of interacting and starburst galaxies which have been
detected in the IRAS survey, but now these objects typically lie at
redshifts 
.
I will discuss two examples of the study of galaxies at very large
redshifts. The first of these results from the study by Windhorst &
Keel (1995) of what they refer to as a young `elliptical' radio galaxy
at a redshift 
. Although a radio galaxy, and so possibly
unusual in it characteristics, it is a relatively modest radio emitter,
about 100 times less luminous radio-wise than the objects I discuss in
Section 7. In their paper, they derive a surface brightness profile for
the galaxy which more or less follows a de Vaucouleurs 
law,
and this is the basis of their claim that it is an elliptical galaxy.
At this meeting, Pascarelle described new observations of the field of
this radio galaxy which, by great good fortune, lies at such a redshift
that the Lyman-
line is redshifted into the narrow F410W filter.
They find evidence for 18 Lyman-
objects at this redshift, all
of them with luminosities between about 0.1 and 1 
. All of these
objects seem to be compact and again the sum of their brightness
distributions seems to follow the 
law. They suggest that this
is evidence for the early formation of the bulges of galaxies.
Figure: (a). The spectrum of a starburst of duration 12Gyr as
observed at different ages (White 1989, from computations by G.
Bruzual). (b) Illustrating how three colour photometry in the U
, G
and R
wavebands can isolate star-forming galaxies at large
redshifts. The dashed line shows the spectrum of a star-forming galaxy
at a redshift z = 3 (Macchetto & Giavalisco 1995).
The largest redshift systems which have been identified as young
star-forming galaxies have been discovered by searching for the
redshifted Lyman limit by multi-colour photometry. The technique is
similar to that described by Lilly & Cowie (1987) and refined for the
detection of `Lyman-limit galaxies' by Steidel & Hamilton (1992,1993).
The predicted spectrum of a starburst galaxy is illustrated in Figure 5(a)
in which it can be seen that, as the starburst ages, the spectrum remains
of the same characteristic form, namely, it is flat from the Lyman limit
at 91.2 nm to longer wavelengths with an abrupt cutoff at 
nm. At a redshift z = 3, the Lyman limit is shifted to 400 nm
and so the characteristic signature of these objects is that roughly
equal intensities are observed in the G and R
wavebands but
the intensity in the ultraviolet waveband is very low as illustrated in
Figure 5(b). The story began with the successful attempt to identify the
large redshift absorption systems present in the background quasar QSO
0000-262, which has an emission redshift 
(Steidel & Hamilton
1992,1993). In this field, Macchetto et al. (1993) identified a
`Lyman-
radio quiet galaxy' at a redshift 
. Searches
in four other QSO fields are described by Steidel et al. (1995).
Macchetto & Giavalisco (1995) and Steidel et al. (1996) have
described further observations of these fields. Macchetto &
Giavalisco (1995) described the application of this multi-colour
technique to the field containing the galaxy at redshift 
and
several objects with the signature of star-forming galaxies were
found. At this meeting, Giavalisco described the exciting result that
spectroscopy with the Keck 10-m telescope has confirmed that these
objects are indeed galaxies at redshift 
. These galaxies
have been imaged by the WFPC2 and, when the images of the galaxies are
summed, they are found to follow the standard de Vaucouleurs 
dependence of surface brightness upon radius of elliptical galaxies.
Figure: The central portion of the field of the cluster Abell
2218 as observed through the F702W filter of the WFPC2. (Kneib et al.
1996).
Steidel et al. (1996) have obtained the spectra of 24 candidate
star-forming galaxies selected in both the quasar fields and in random
regions of sky and have had great success in measuring redshifts for
these with the Keck 10-m telescope. Seventeen of the objects have
redshifts in the interval 
. They find the important
result that the co-moving space density of these star-forming galaxies in
the redshift interval 
is about half that of luminous
galaxies with 
at the present epoch. The inferred velocity
dispersions within the galaxies suggest that they are indeed massive
galaxies. The star formation rates correspond to about about 
yr
, similar to the star formation rates
per galaxy found by Cowie et al. (1995). Steidel et al.
infer that they have discovered the formation of the spheroidal
components of the progenitors of massive galaxies --- massive galaxy
formation was certainly well underway by a redshift of 3.
As was emphasised in Section 2, normal galaxies at large redshifts can only be readily observed if their surface brightnesses are enhanced. One clever way of achieving this is to use the gravitational lensing of rich clusters of galaxies to magnify the flux densities of distant background galaxies. A remarkable example of how this can be done is described by Kneib et al. (1996) who have studied in detail the gravitational lensing of distant galaxies by the mass in the central regions of the cluster Abell 2218. Figure 6 shows their beautiful image of the central regions of the cluster with the remarkable gravitationally lensed images of distant background objects. There are several important features of the lensed background objects which enable the mass distribution in the lens to be determined with considerable precision. In particular, the observation of multiple images of the same background object enable the location of the critical lines and the detailed mass distribution to be determined rather precisely.
Once the mass distribution of the lens has been determined, the images of other distant background objects are expected to be sheared and distorted at different projected distances from the centre of the lens, in ways which depend upon their redshifts. Kneib et al. (1996) describe in detail the procedures involved in making these redshift estimates and also the uncertainties involved, among these being knowledge of the intrinsic shapes of the galaxies. Consequently, only statistical estimates of the redshifts of the galaxies can be made. Using these procedures, Kneib et al. have estimated the redshifts of the faint elongated images in the field of the cluster Abell 2188 with the results shown in Figure 7. It is apparent that this technique provides a means of studying the properties of galaxies at much larger redshifts than would have been possible without the intervention of the gravitational lens. Just after the meeting in Paris, Ellis (1996) reported at the ICGC95 meeting in Pune, India, that redshifts have now been obtained for a number of these faint galaxy images and the spectroscopic redshifts are in remarkable agreement, in general, with the estimates based upon the gravitational shearing of their images, confirming the great potential of this technique for studying galaxy evolution at large redshifts.
Figure: The estimated redshifts of the faint elongated objects observed
in the field of the cluster Abell 2218. The lines show the expected
mean redshift distribution for the lensed objects. (Kneib et al.
1996).
Finally, let me describe some of our recent HST observations of radio
galaxies at redshifts 
. The programme consists of WFPC2 imaging
observations of a complete sample of 3CR radio galaxies in the redshift
interval 
. These radio galaxies are of special interest
because it is known that they exhibit the strong cosmological
evolutionary trends observed in the radio source and quasar populations
as a whole. It is also known that, in the majority of cases, the optical
structures are aligned with the radio axes of these double radio
sources. We have observed all these radio galaxies with the HST in
wavebands corresponding more or less to rest-wavelength U and B
wavebands. These observations have been supplemented by infrared
observations at 2.2 
m taken with UKIRT which have angular
resolution of about 1 arcsec and by 8.4 GHz VLA observations with angular
resolution 0.18 arcsec. The infrared observations provide images of
the old stellar populations in these galaxies and they all resemble
standard giant elliptical galaxies.
Figure: HST and UKIRT images of the radio galaxies 3C 266, 368,
324, 280 and 65 with the VLA radio contours superimposed. The images
are drawn to the same physical scale (Best, Longair & Röttgering
1996a).
Figure: HST and UKIRT images of the radio galaxies 3C 267, 252
and 356 with the VLA radio contours superimposed. The images are drawn
to the same physical scale (Best, Longair & Röttgering
1996a).
The high resolution HST images are dramatically different. Virtually all
the images show emission regions aligned with the jets which are assumed
to be powering the hot-spots in the outer radio lobes. Perhaps the most
striking result is the comparison of the maps of all the radio galaxies
in our sample in the redshift interval 
(Best, Longair
& Röttgering 1996a). Since these are all 3CR radio galaxies, the
radio sources have the same intrinsic luminosities, implying that the
jet luminosities are the same for all of them. Figures 8 and 9 show
a montage of these eight radio galaxies in order of increasing separation
between the components of the double radio sources. It can be seen that
the most remarkable structures are associated with the smaller double
radio sources. The optical emission regions associated with the radio
galaxies 3C 266, 368, 324 and 280 are all aligned along the axis of the
double radio sources. Comparison of the optical and infrared images shows
that the optical structures are on more or less the same physical scale as
the host galaxy. As the sizes of the double source increase, the optical
emission regions become less prominent and, although there is still some
alignment with the radio axis, the structures are on a smaller physical
scale. According to the theory of double radio sources, the large sources
are older than the smaller sources and what is of particular interest in
these cases is that, for some of them, synchrotron ages are available.
These arguments suggest that the radio sources associated with 3C 266
and 280 are a few million years old (Liu, Pooley & Riley 1992).
These observations suggest that the strongly-aligned optical
structures are short-lived phenomena, which are stimulated by passage
of the radio jet. As we have pointed out already (Longair, Best &
Röttgering 1995), it seems that no single theory of these alignments
can account for all the observations. The observation of polarised
optical emission from some of these sources suggests that scattering
of light from an obscured quasar must play some role in accounting
for the emission from these optical structures. It seems, however,
that the primary cause of the structures must be the interaction of
the radio jet with cool interstellar clouds within the parent galaxy.
Exactly how these structures are formed is unclear. One possibility
is that the structures are associated with jet-induced star-formation.
It would then be possible to account rather naturally for the change in
structure with increasing physical size. The lifetime of the newly formed
stars and associated HII regions would amount to only about 
years.
After this time, the luminosities of the star-forming regions would decay
and the star clusters relax within the potential of the parent giant
elliptical galaxy. The polarisation of the light would be attributed
to the scattering of the light of an obscured quasar by the dust or gas
associated with the star-forming regions. A problem with this picture is
that there have so far been no reports of young stars in the spectra of
the emission regions. Alternatively, the alignments may be due to the
illumination of pre-existing dust and gas clouds by a central obscured
quasar. There remains the problem of accounting for the presence of
large amounts of cool gas and dust within the body of the parent galaxy.
Figure: (a) The HST image of the radio galaxy 3C 34 superimposed upon
which are the VLA radio contours of the double radio source structure.
(b) and (c) Images of the structure of the optical `jet' which lies
along the line from the nucleus of the galaxy to the western hot-spot as
observed through the f555W and f785LP filters respectively. (d) J and
(e) K images of the galaxy associated with the optical jet observed with
UKIRT (Best, Longair & Röttgering 1996b)
One remarkable result, which may be evidence for jet-induced star-formation, has been found in the field of the radio galaxy 3C 34, which has a redshift of 0.69. In this source, the radio galaxy is rather diffuse with no prominent structures such as those found in the smaller double sources, as can be seen in Figure 10. However, along the axis from the nucleus of the radio galaxy to the brightest western hot-spot, there is a remarkable linear feature which passes close to the galaxy labelled (a). These structures are shown in more detail in Figure 10(b-e). The linear feature is bluer than other objects in the field. Our interpretation of this feature is that they may represent the interaction of the radio jet with the interstellar gas in a galaxy which just happens by accident to lie in the path of the jet.
Abraham, R.G., van den Bergh, S., Glazebrook, K., Ellis, R.S., Santiago, B.X., Surma, P. & Griffiths, R.E. 1996, MNRAS, in press
Best, P.N., Longair, M.S. & Röttgering, H.J.A. 1996a, MNRAS, in press
Best, P.N., Longair, M.S. & Röttgering, H.J.A. 1996b, MNRAS, in preparation
Cowie, L.L. 1988, in The Post-recombination Universe, eds. N. Kaiser & A.N. Lasenby, 1. Dordrecht: Kluwer Academic Publishers
Cowie, L.L., Hu, E.M. & Songaila, A. (1995), Nature, 337, 603
Driver, S.P., Windhorst, R.A. & Griffiths, S.P. 1995, ApJ, 453, 48
Driver, S.P., Windhorst, R.A., Ostrander, E.J., Keel, W.C., Griffiths, R.E. & Ratnatunga, K.U. 1995, ApJ, 449, L23
Ellis, R.S. 1996, in International Conference on Gravitation and Cosmology, ICGC95, Pune, India, ed. T. Padmanabhan, in press
Giavalisco, M, Livio, M., Bohlin, R.C., Macchetto, F.D. & Stecher, T.P. 1996, preprint
Glazebrook, K., Ellis, R.S., Santiago, B. & Griffiths, R. 1995a, MNRAS, 275, L19
Glazebrook, K., Ellis, R., Colless, M., Broadhurst, T., Allington-Smith, J. & Tanvir, N. 1995b, MNRAS, 273, 157
Kneib, J-P, Ellis, R.S., Smail, I, Couch, W.J. & Sharples, R.M. 1996, MNRAS, in press
Gunn, J.E 1979, in Scientific Research with the Space Telescope, IAU Colloquium No. 54, eds. M.S. Longair & J. Warner, 313, NASA CP-2111
Le Fèvre, O., Crampton, D., Lilly, S.J., Hammer, F. & Tresse, L. 1995, ApJ, 455, 60
Lilly, S.J. & Cowie, L.L. 1987, in Infrared Astronomy with Arrays, eds. C.G. Wynn-Williams & E.E. Becklin, 473. Honolulu: Institute for Astronomy, University of Hawaii Publications
Liu, R., Pooley, G.G. & Riley, J.M. 1992, MNRAS, 257, 545
Longair, M.S., Best, P.N. & Röttgering, H.J.A. 1995, MNRAS, 275, L47
Macchetto, F.D. & Giavalisco, M. 1995, ESO Messenger, September 1995, 14
Macchetto, F.D., Lipari, S., Giavalisco, M., Turnshek, D.A. & Sparks, W.B. 1993, ApJ, 404, 511
Neuschaefer, L.W., Ratnatunga, K.U., Griffiths, R.E & Casertano, S. 1996, ApJ, in press
Schade, D., Lilly, S.J., Crampton, D., Hammer, F., Le Fèvre, O. & Tresse, L. 1995, ApJ, 451, L1
Steidel, C.C., Giavalisco, M., Pettini, M., Dickinson, M. & Adelberger, K.L. 1996, AJ, in press
Steidel, C.C. & Hamilton, D. 1992, AJ, 104, 941
Steidel, C.C. & Hamilton, D. 1993, AJ, 105, 2017
White, S.M. 1989, in The Epoch of Galaxy Formation eds. C.S. Frenk, R.S. Ellis, T. Shanks, A.F. Heavens & J.A. Peacock, 1. Dordrecht: Kluwer Academic Publishers
Windhorst, R.A., Fomalont, E.B., Kellermann, K.I., Partridge, R.B., Richards, E., Franklin, B.E., Pascarelle, S.M. & Griffiths, R.E. 1995, ApJ, 375, 471
Windhorst, R.A. & Keel, W.C. 1996, ApJ, in press